Instructions to use moonbox-dev/flux2_klein with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use moonbox-dev/flux2_klein with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download moonbox-dev/flux2_klein --local-dir flux2_klein
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
FLUX.2 [klein] 4B for Moonbox
The files the Moonbox desktop app downloads for these models:
- FLUX.2 klein 4B (Black Forest Labs): Fast, sharp pictures and edits from up to three references.
They are FLUX.2 [klein] 4B converted to MLX for Apple silicon and quantized to 8-bit, with its text encoder, tokenizer and VAE. Moonbox checks every file against a signed list before using it. You do not need this repository to use Moonbox, and it is not an official release of the original models.
What is here
| Path | What | From |
|---|---|---|
outputs/export/klein-4b-q8/transformer/, outputs/export/klein-4b-q8/vae/, outputs/export/klein-4b-q8/model_index.json, outputs/export/klein-4b-q8/quant.json |
FLUX.2 [klein] 4B (distilled transformer + FLUX.2 VAE) | https://huggingface.co/black-forest-labs/FLUX.2-klein-4B |
outputs/export/klein-4b-q8/text_encoder/, outputs/export/klein-4b-q8/tokenizer/ |
Qwen3-4B text encoder + tokenizer (as shipped inside FLUX.2-klein-4B) | https://huggingface.co/black-forest-labs/FLUX.2-klein-4B/tree/main/text_encoder (upstream Qwen/Qwen3-4B) |
models/hf/FLUX.2-klein-4B/ |
the text encoder, tokenizer, VAE and configs as published, for the Windows engine | https://huggingface.co/black-forest-labs/FLUX.2-klein-4B |
models/torch/ |
the transformer (Q6_K, Q8_0 and W8A8 exports) and an 8-bit text encoder, for the Windows engine | https://huggingface.co/black-forest-labs/FLUX.2-klein-4B |
Every part was converted, quantized or repackaged from its source unless its row says "as published": treat none of these files as the original release.
Licences
FLUX and Black Forest Labs are trademarks of Black Forest Labs. Their use here names the model these files were made from and implies no sponsorship or endorsement.
| Part | Licence | Notice |
|---|---|---|
| FLUX.2 [klein] 4B (distilled transformer + FLUX.2 VAE) | Apache License 2.0 | Black Forest Labs. Use of Black Forest Labs and FLUX trademarks or logos in modified versions of this project must not cause confusion or imply sponsorship or endorsement. |
| Qwen3-4B text encoder + tokenizer (as shipped inside FLUX.2-klein-4B) | Apache License 2.0 | Copyright 2024 Alibaba Cloud (Qwen3-4B LICENSE) |
| the text encoder, tokenizer, VAE and configs as published, for the Windows engine | Apache License 2.0 | the upstream LICENSE.md is in the folder |
| the transformer (Q6_K, Q8_0 and W8A8 exports) and an 8-bit text encoder, for the Windows engine | Apache License 2.0 |
The full text of each licence is in LICENSE. The models are the work of their authors, linked above. Moonbox converted and packaged them and claims no rights in them. Use is subject to each licence above.
Quantized
Model tree for moonbox-dev/flux2_klein
Base model
black-forest-labs/FLUX.2-klein-4B